Package 4 · Services

AI architectures and automation

For organisations that must run AI in their own network – without cloud and without outside access.

Run AI in your own network and automate processes – compliant with your rules.

How you know you need this

AI is required, but your data must not leave the building – because of VS-NfD, supervisory requirements or your own rules.

Business units already use public AI services because there is no alternative. You want to control that without banning the use.

Recurring operational work consumes time, and every manual intervention is a potential source of error.

Clarifying the legal framework and requirements before hardware is bought.
Decision paper for hardware and platform: what carries which model size, what operations cost.
Build and integration in your own network, including connection to existing identities and logging.
Assessment of the tools to be used in-house – with a view to data egress and licensing.
Automation of recurring processes as code (Terraform, Ansible, pipelines) so environments are reproducible.
Operations handover: processes, monitoring, incidents, model updates.

Build usually six to twelve weeks, then operations run. Automation is added step by step.

A test environment with the intended hardware, a contact for network and identities, a statement on which data is permitted.

Procurement of hardware and licences. Recommendation and decision paper are included, purchasing stays with you – that keeps interests separate.

Open models that can run in your own network. Size follows task and hardware – for many cases a small model is sufficient.

Essentially the hardware you would procure anyway. Ongoing licence costs for external services disappear because processing stays in-house.

Yes. The architecture is chosen so that a larger model or further tasks can be added without rebuilding everything.

In the first call we clarify whether the package fits your situation. Then you get two proposed dates and a clear statement on effort.

Architecture review – For organisations whose IT grew over years and where nobody has the full picture any more.
Project and sub-project management – For undertakings that need one accountable person – not another supplier.
Interim: lead architect / platform operations lead – For vacancies that operations cannot wait for.
IT audit, DORA and ICT risk management – For financial entities and regulated organisations that must be audit-ready.
Training and seminars – For teams that should be able to do it themselves afterwards.